Security Data Works

Public production architecture teardown

Microsoft Sentinel on Azure

Azure-native managed SIEM built on Log Analytics Workspace (columnar Kusto storage) with KQL as the query language. Schema-on-read at the storage layer; zero-config ingest for Azure AD, Defender, Office 365; Logic Apps for response automation. The architecture is documented; the cost model is what bends the decision at scale.

$2.30/GB

Pay-as-you-go ingestion rate for the first 100 GB/day (Q1 2026 published pricing). Commitment tiers discount at higher volume. Cost includes ingestion, 90-day retention, and amortized query compute — schema-on-read query overhead is bundled into the per-GB rate.

The pipeline

  1. Sources

    Azure-native + 3rd party

    Azure AD · Defender · O365 · CrowdStrike · Palo Alto

  2. Ingest

    Data connectors

    Diagnostic Settings · API polling · syslog/CEF

  3. Store

    Log Analytics Workspace

    Kusto columnar; 30–730 day retention

  4. Detect

    Analytics rules (KQL)

    Scheduled queries; 5-min default cadence

  5. Respond

    Logic Apps playbooks

    Email · ServiceNow · isolate endpoint · block IP

What composes, what’s brittle

  • Query engine. Kusto — same engine as Azure Data Explorer.
  • Schema-on-read. Nested JSON re-parsed on every query; cost amortized into $2.30/GB.
  • E5 bundling. M365 E5 includes a 5 MB/user/day Sentinel data grant on Microsoft sources.
  • Best fit. 80%+ Azure workloads, < 100 GB/day, no data engineering team.
  • What's brittle. Cost at > 500 GB/day; KQL non-portable; 90-day retention wall.
  • Migration tool. Microsoft-published SPL → KQL translator (~70–80% automated).

Sources: Microsoft Azure pricing page (Sentinel, Q1 2026) · Microsoft Learn "Log Analytics workspace overview" · Microsoft Learn migration guidance · Azure-Sentinel/Tools/RulesTranslators GitHub

See how the pattern lands on your workload.

The matrix scoring that justified each reference architecture's tool choices is the paid deliverable. The benchmark behind it is public — reproduce it on your own workload, then book a call to scope the work.